Attention-based layered neural network architecture for explainable and high-performance ai processing

Abstract

A system for attention-based layered neural network classification is provided. The system comprises: a sequence of layered neural networks; and a controller configured for controlling data routed through the sequence of layered neural networks, the controller configured to: receive interaction data comprising data features, wherein the data features are distinct characteristics of the interaction data; input data features into the sequence of layered neural networks, wherein each sequential layer of the sequence of layered neural networks comprises a heightened rigor level for at least one of the data features; calculate a relevance score output for at least one of the data features at each layer of the sequence of layered neural networks; and integrate the relevance score output from each layer of the sequence of layered neural networks to generate a total relevance score output.

Description

BACKGROUND[0001]As artificial intelligence techniques, such as machine learning, have become more commonly used in decision making processes, explainability for these decisions has become increasingly important for reliability and proper use of such systems. In recent years, a wide range of decisioning systems have started incorporating artificial intelligence systems in their processes. Non-limiting examples of these systems include airport or public transportation security systems, misappropriation detection systems, and medical diagnostic systems. Explainability in artificial intelligence decisioning has also attracted significant attention from regulators, as critical decisions are increasingly made by such systems. As such, regulations are beginning to mandate requirements for explainability of the underlying decisions made by artificial intelligence systems.[0002]That said, explainability and interpretability remain as one of the grand challenges in artificial intelligence. Va...

AI Technical Summary

Benefits of technology

The invention is a system for classifying data based on attention-based layered neural networks. The system includes a sequence of layered neural networks and a controller for controlling data routing through the networks. The controller receives interaction data and inputs the data into the first layer of the network. Each layer of the network has a higher rigor level for certain data features. The system calculates a relevance score output for each data feature at each layer and integrates the scores to generate a total relevance score output. The system can be used for identifying abnormal or unauthorized patterns in the data.

Problems solved by technology

That said, explainability and interpretability remain as one of the grand challenges in artificial intelligence.

While most entities and users have a strong interest in responsible and explainable decision making with the help of artificial intelligence, current tools and technologies to enable these goals are very limited at this point.

Images